The portion of the public network from a central office to an end user's location is called the access network or “last mile.” The access network connects the end user to a backbone or core network via the central office. To keep pace with increasing voice, data, and video traffic, network operators have, in many areas, upgraded existing access networks by deploying optical fibers deeper into the last mile to shorten the lengths of existing copper and coaxial networks. Among different competing optical network technologies, passive optical networks (PONs) have been one of the favored choices for these next-generation access networks. With the large bandwidth of optical fibers, PONs can accommodate bandwidth-intensive voice, data, and video services.
FIG. 1 illustrates an exemplary PON 100 that communicatively couples a central office 110 to a single family unit (SFU) 120 and a multi-dwelling unit (MDU) 130 (i.e., a structure housing two or more residential or business units). Transmissions within PON 100 are specifically performed between an optical line terminal (OLT) at central office 110 and optical network units (ONUs) at SFU 120 and MDU 130 over optical fibers that span the distance between them. The OLT at central office 110 couples PON 100 at its end to a backbone or core network (not shown), which can be an external network belonging to, for example, an Internet service provider (ISP) or a local exchange carrier. In addition, the ONUs at SFU 120 and MDU 130 further couple PON 100 at their ends to home or business networks (also not shown). This network structure allows end user devices coupled to the home or business networks within SFU 120 and MDU 130 to send data to, and receive data from, the backbone or core network over PON 100.
The portion of PON 100 closest to central office 110 is commonly referred to as the feeder area 150. This area includes one or more feeder cables that each has multiple fibers. Passive optical splitters/combiners 140 are used to split the individual fibers of the feeder cables into multiple distribution fibers that fall within the second portion of PON 100, which is commonly referred to as the distribution area 160. The distribution fibers are then further split by additional passive optical splitters/combiners 140 into multiple drop fibers that extend to SFU 120 and MDU 130. The drop fibers fall within the third and final portion of PON 100, which is commonly referred to as the drop area 170.
It should be noted that PON 100 illustrates only one exemplary PON and fiber distribution topology (i.e., a tree topology) and that other point-to-multipoint fiber distribution topologies, such as ring and mesh topologies, are possible.
In prior access networks, distribution area 160 and/or drop area 170 were deployed using copper and coaxial cables. By extending fiber cables deeper into the access network, all the way to the home, building, or curb, for example, PON 100 can accommodate bandwidth-intensive voice, data, and video services that these prior access networks could not handle. As illustrated in FIG. 1, the only remaining portion of the network between central office 110 and an end user's device at SFU 120 and MDU 130 that potentially is not optically connected is within the local area networks at these locations (i.e., within metallic area 180). Over such short copper and/or coaxial wiring distances, current local area network technology generally provides adequate bandwidth.
During operation of the access network illustrated in FIG. 1, signals sent down stream over the three portions of PON 100 by the OLT at central office 110 are split by passive optical splitters/combiners 140 and are received by the ONUs at SFU 120 and MDU 130. Conversely, signals sent upstream over the three portions of PON 100 by the ONUs at SFU 120 and MDU 130 are combined by passive optical splitters/combiners 140 and are received by the OLT at central office 110. To prevent collisions in the upstream direction and to share the upstream capacity of PON 100 fairly, the OLT at central office 110 and the ONUs at SFU 120 and MDU 130 implement some form of arbitration mechanism.
For example, many PONs implement a non-contention based media access scheme that grants each ONU access to the shared medium for a limited interval of time for transmitting data upstream. This limited interval of time is commonly referred to as a time slot. FIG. 2 illustrates an example of data being sent upstream over a PON in accordance with such a non-contention based media access scheme. In FIG. 2, each ONU 1 through N is synchronized to a common time reference and is allocated a timeslot for transmitting one or more packets of data upstream to the OLT. More specifically, each ONU 1 through N buffers packets received from an attached end user (or users) and bursts one or more of the buffered packets upstream to the OLT when its assigned timeslot arrives. For example, ONU 1 receives two packets of data from attached user 1, buffers the two packets of data, and bursts the two packets upstream during a first timeslot assigned to ONU 1. ONU 2 receives a single packet of data from attached user 2, buffers the packet of data, and bursts the single packet upstream during a second timeslot assigned to ONU 2. As can be seen from FIG. 2, the time slots are assigned to the ONUs such that they do not overlap in time, thereby preventing upstream collisions.
Beyond simply assigning time slots such that they do not overlap in time, the exact method of when and how much capacity is granted to a particular ONU in such a non-contention based media access scheme can greatly affect the performance of the PON. In most conventional PONs, each ONU is either assigned a static time slot of fixed capacity within a repeating interval (i.e., in accordance with a static TDMA scheme) or are dynamically assigned timeslots of varying capacities based on the instantaneous amount of data buffered by the ONUs (i.e., in accordance with a dynamic bandwidth allocation (DBA) scheme).
Static TDMA schemes are advantageous because ONUs are guaranteed a predetermined, fixed amount of upstream bandwidth. However, the inflexibility of the amount of upstream bandwidth assigned to each ONU using static TDMA often leads to inefficient use of upstream bandwidth—ONUs with little or no data to transmit upstream are still allocated their fixed share of bandwidth even though other ONUs may be waiting to send large amounts of data upstream. In other words, static TDMA does not allow unused upstream bandwidth to be transferred to ONUs that can make use of the excess bandwidth. DBA schemes, on the other hand, allow upstream bandwidth to be more efficiently allocated to, and shared by, ONUs based on need and other factors.
In a PON implementing a DBA scheme, the OLT is responsible for allocating upstream grants (or timeslots) to each ONU. An ONU defers its upstream transmissions until it receives a grant from the OLT. In order to receive a grant, an ONU generates and transmits an upstream message to the OLT called a REPORT message that informs the OLT of its respective upstream queue status. The OLT uses this information sent from each ONU requesting upstream bandwidth to generate and transmit GATE messages to those ONUs. Each GATE message allocates an upstream transmission grant to an ONU based on its upstream bandwidth needs and the upstream bandwidth needs of other ONUs.
In Ethernet PONs (EPONs), the packets buffered by the upstream queues of the ONUs are Ethernet frames, which generally cannot be fragmented. Therefore, if an OLT allocates an upstream transmission grant to an ONU that is less than the total amount of data buffered by its upstream queue, or allocates an upstream transmission grant that does not fall on a boundary of an Ethernet frame in its upstream queue, some of the upstream bandwidth may go to waste. FIG. 3 provides a specific illustration of this, where an ONU is provided with an upstream transmission grant 310 that is less than the total amount of data buffered by its upstream queue 320 and does not fall on a boundary of an Ethernet frame in its upstream queue 320. The upstream transmission grant 310 is sufficient to transmit all of frames 1-3 in upstream queue 320, but is not sufficient to transmit all of frame 4. Because frames in upstream queue 320 cannot be fragmented, there is an upstream bandwidth loss 330.
The embodiments of the present disclosure will be described with reference to the accompanying drawings. The drawing in which an element first appears is typically indicated by the leftmost digit(s) in the corresponding reference number.